run_demo_loop — Interactive REPL for Agents

A tiny terminal REPL for agents. Useful for smoke-testing, demos, and local exploration without writing a wrapper script.

Signature

async def run_demo_loop(
    agent: Agent[Any],
    *,
    stream: bool = True,
    context: Optional[Any] = None,
    max_turns: int = DEFAULT_MAX_TURNS,
    run_config: Optional[RunConfig] = None,
    hooks: Optional[RunHooks[Any]] = None,
) -> None

All parameters are forwarded to Runner.arun(). The loop preserves conversation history across turns by calling RunResult.to_input_list() and feeding the result back in as the next user prompt.

Behavior

• Reads input from stdin with the prompt >
• Streams tokens live when stream=True (default)
• Exits cleanly on exit, quit (case-insensitive), EOF (Ctrl-D), or SIGINT (Ctrl-C)
• Skips empty input without making an LLM call
• Tracks handoffs — the currently active agent after the last turn becomes the prompt for the next turn

Usage

import asyncio
from augments.adk.agents.agent import Agent
from augments.adk.run.demo import run_demo_loop
 
agent = Agent(
    name="assistant",
    system_prompt="You are a helpful assistant. Keep answers concise.",
)
 
asyncio.run(run_demo_loop(agent))

Then, in the terminal:

> What's the capital of France?
Paris.
> And of Portugal?
Lisbon.
> exit

Non-streaming mode

Passing stream=False buffers each turn's output and prints it once the response is complete — handy for structured-output agents whose intermediate tokens aren't meaningful text.

asyncio.run(run_demo_loop(agent, stream=False))

See also

• examples/agent_patterns/demo_loop.py — runnable example
• src/augments/adk/run/demo.py — implementation
• tests/unit/run/test_demo_loop.py — behavior tests